What you think about the state of the race for president depends upon what sources you are looking at. 

Rasmussen Reports is bullish on Trump, and the Atlas Intel polls--they were the most accurate pollster over the past two election cycles--released some polls last month that showed a likely Trump electoral college blowout. 

The polling averages look good for Trump, although they show a razor-thin difference between the candidates. 

But when you zoom out a bit, a lot of the polls look good for Harris, too, showing her having made up ground and having leads in important swing states. As always, it comes down to Pennsylvania, and Harris may have a razor-thin lead in the polls. 

In other words, if you are relying exclusively on the polls, you can find a good case for either Trump or Harris, and perhaps even feel sure of your prediction. 

Both Nate Silver and ABC's 538 show Harris with a lead in the odds of victory, if not a huge one. Both of them have the odds of a Harris victory at 55%, which really isn't that huge a lead. 

The betting at Polymarket is a complete inversion of that, though. 

Based on their predictions of who wins each state, bettors are giving Trump a 9-point edge in odds of victory. The margins state by state are relatively thin--for instance, the smart money only gives Trump a 56% chance of winning Pennsylvania--but add up all the bets, and the best guess of the betting markets right now is that Trump likely takes the election by a small margin. 

What's truly fascinating, though, is not the current guesses about who wins in a month, but rather how sentiment has dramatically changed over the past few days. Here is the chart covering the past week. 

That is a HUGE swing in sentiment, going from a tie in the odds to Trump up by 9. It all basically happened over the weekend.

At first, I thought somebody was manipulating the market, but as you can see, there is $1.4 billion in the kitty, a decent-sized market to manipulate like this, and the Democrats have more money to blow--and they certainly aren't selling Kamala Harris short. 

On the other hand, most betting markets give Harris a slight edge too small to count. 

It is certainly possible that the market is being manipulated, but there are alternative explanations. The one I like is very basic: a belief that, as in 2016 and 2020, the pollsters are undercounting Trump voters by a large margin. If you compare the Biden 2020 numbers to Harris' 2024, her polling looks terrible. 

Trump barely lost the 2020 election, and he is polling 7.6% better this time around compared to that election, and over 4 points better in the battlegrounds. Unless the pollsters have figured out how to poll Trump supporters better than in 2016 and 2020, Harris is toast. Even if their errors are cut by 75%, she is going to lose.

All the prediction models you read assume that the margin of error is actually the margin of error. But in the past, the MOE on Trump is 2-3 times greater than advertised due to "shy" Trump voters. That could be different this time, but undercount has been consistent so far. If there is even a modest Trump undercount, the election looks very different from the Silver/538 models. 

The models rely on hard data, but that data is suspect. Disbelieving the models rely on trends continuing from past elections, and they may not this year. 

Pick your poison. My bet, without real money on the line, is that Trump is probably doing a bit better than the polls show. And if so, that means he wins--if things are fair.